aibl Research: The loud and clear signals of Shadow AI

In keeping with our lead article, we went into the data from aibl’s Mid-market AI survey and dug into unapproved AI.

Any reasonable, governance-minded company wants to minimise ‘shadow AI’, and our research supports that view. Leaders who report higher levels of unapproved AI also report higher rates of AI-related failures, across every dimension we measured.

The straightforward reading is that ungoverned tools mean no vetting, no data controls, no testing standards and no accountability when output is wrong. More unsanctioned use produces more incidents, including bad outputs reaching customers, data leakage, compliance violations, duplicated or conflicting automations. An IBM report finding that one-in-five breaches involved shadow AI appears to support this view.

But we think the real story is about leadership, governance, and providing the right tools. Shadow AI is a demand signal. It tells you where the gaps are. What the data is really asking for is C-suite coordination, clear oversight, and the approved tools to fill those gaps.

More next week on what good governance looks like in practice.